DEFT: Data Efficient Fine-Tuning for Pre-Trained Language Models via Unsupervised Core-Set Selection

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Das, Devleena, Khetan, Vivek
Natura: Preprint
Pubblicazione: 2023
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929383843823616
author Das, Devleena
Khetan, Vivek
author_facet Das, Devleena
Khetan, Vivek
contents Recent advances have led to the availability of many pre-trained language models (PLMs); however, a question that remains is how much data is truly needed to fine-tune PLMs for downstream tasks? In this work, we introduce DEFT-UCS, a data-efficient fine-tuning framework that leverages unsupervised core-set selection to identify a smaller, representative dataset that reduces the amount of data needed to fine-tune PLMs for downstream tasks. We examine the efficacy of DEFT-UCS in the context of text-editing LMs, and compare to the state-of-the art text-editing model, CoEDIT. Our results demonstrate that DEFT-UCS models are just as accurate as CoEDIT, across eight different datasets consisting of six different editing tasks, while finetuned on 70% less data.
format Preprint
id arxiv_https___arxiv_org_abs_2310_16776
institution arXiv
publishDate 2023
record_format arxiv
spellingShingle DEFT: Data Efficient Fine-Tuning for Pre-Trained Language Models via Unsupervised Core-Set Selection
Das, Devleena
Khetan, Vivek
Computation and Language
Artificial Intelligence
Recent advances have led to the availability of many pre-trained language models (PLMs); however, a question that remains is how much data is truly needed to fine-tune PLMs for downstream tasks? In this work, we introduce DEFT-UCS, a data-efficient fine-tuning framework that leverages unsupervised core-set selection to identify a smaller, representative dataset that reduces the amount of data needed to fine-tune PLMs for downstream tasks. We examine the efficacy of DEFT-UCS in the context of text-editing LMs, and compare to the state-of-the art text-editing model, CoEDIT. Our results demonstrate that DEFT-UCS models are just as accurate as CoEDIT, across eight different datasets consisting of six different editing tasks, while finetuned on 70% less data.
title DEFT: Data Efficient Fine-Tuning for Pre-Trained Language Models via Unsupervised Core-Set Selection
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2310.16776